[Paper Review] CovMUNET: A Multiple Loss Approach towards Detection of COVID-19 from Chest X-ray

Ali Sayyed, Dipayan Saha+1 more|arXiv (Cornell University)|Jul 28, 2020

COVID-19 diagnosis using AI4 citations

TL;DR

CovMUNET proposes a deep learning model using a multiple loss function to detect COVID-19 from chest X-ray images, achieving 96.97% accuracy in 3-class classification (COVID-19 vs. normal vs. pneumonia) and 99.41% in 2-class classification (COVID vs. non-COVID), demonstrating strong performance and robustness in identifying abnormalities in CXR scans.

ABSTRACT

The recent outbreak of COVID-19 has halted the whole world, bringing a devastating effect on public health, global economy, and educational systems. As the vaccine of the virus is still not available, the most effective way to combat the virus is testing and social distancing. Among all other detection techniques, the Chest X-ray (CXR) based method can be a good solution for its simplicity, rapidity, cost, efficiency, and accessibility. In this paper, we propose CovMUNET, which is a multiple loss deep neural network approach to detect COVID-19 cases from CXR images. Extensive experiments are performed to ensure the robustness of the proposed algorithm and the performance is evaluated in terms of precision, recall, accuracy, and F1-score. The proposed method outperforms the state-of-the-art approaches with an accuracy of 96.97% for 3-class classification (COVID-19 vs normal vs pneumonia) and 99.41% for 2-class classification (COVID vs non-COVID). The proposed neural architecture also successfully detects the abnormality in CXR images.

Motivation & Objective

• To develop a deep learning model that improves the accuracy of COVID-19 detection from chest X-ray images.
• To address the challenge of limited and imbalanced medical imaging data for COVID-19 by leveraging a multiple loss training strategy.
• To enhance model generalization and robustness in distinguishing between COVID-19, normal, and pneumonia cases in CXR scans.
• To enable early and cost-effective detection of COVID-19 using accessible imaging modalities.
• To support clinical decision-making by identifying abnormalities in chest X-rays with high precision and recall.

Proposed method

• The model employs a multiple loss function combining cross-entropy and focal loss to improve classification performance on imbalanced datasets.
• A custom convolutional neural network architecture is designed to extract hierarchical features from chest X-ray images.
• The network is trained end-to-end using a combination of classification losses tailored for multi-class and binary classification tasks.
• Data augmentation techniques are applied to increase training data diversity and improve model generalization.
• The model is evaluated using standard metrics including precision, recall, F1-score, and accuracy across multiple experimental settings.
• The architecture is optimized to detect not only COVID-19 but also general abnormalities in chest X-rays.

Experimental results

Research questions

• RQ1 Can a multiple loss function improve the detection accuracy of COVID-19 in chest X-ray images compared to single-loss approaches?
• RQ2 How does the proposed model perform in distinguishing between COVID-19, normal, and pneumonia cases in a multi-class setting?
• RQ3 To what extent does the model generalize to unseen chest X-ray data with varying pathologies?
• RQ4 Can the model achieve high precision and recall in binary classification (COVID vs. non-COVID) despite class imbalance?
• RQ5 How effective is the model in detecting general abnormalities in chest X-ray images beyond just COVID-19?

Key findings

• The model achieved 96.97% accuracy in 3-class classification (COVID-19 vs. normal vs. pneumonia), outperforming state-of-the-art methods.
• In binary classification (COVID vs. non-COVID), the model reached 99.41% accuracy, indicating strong discriminative capability.
• The multiple loss strategy significantly improved model robustness and generalization on imbalanced datasets.
• The model demonstrated high precision and recall across all classes, indicating reliable detection performance.
• The architecture successfully detected abnormalities in CXR images beyond just COVID-19, suggesting broader diagnostic utility.
• The results were validated using standard evaluation metrics, confirming the model's effectiveness and reliability.